Neurodegenerative diseases affect the cytoarchitecture of the white matter which hosts neuronal connections. Statistical assessment of its integrity is therefore key for diagnosis, improved therapies and better patient outcome. Diffusion MRI enables visualization of the white matter wiring in-vivo and non-invasively. This talk will focus on comparing specific biomarkers of neurodegenerative diseases provided by diffusion MRI as microstructural parameters evolving along infinite-dimensional curves, a.k.a. streamlines, using functional data analysis. First, we will go into details on the elaboration of the functional data from diffusion MRI including diffusion modeling, tractography algorithms, extraction of bundles of streamlines of interest, a.k.a. tracts, and functional representation of tracts using a $k$-medoid alignment approach and simple outlier detection strategies. Second, we will setup an inferential procedure for comparing the tracts' representative medoids between healthy subjects and patients diagnosed with a specific neurological condition, using intervalwise testing. Specifically, we will show that healthy subjects present statistical differences in their structural tracts, which is consistent with clinical knowledge but problematic for detecting clinically relevant anomalies along tracts because statistical differences could reflect solely normal healthy variations. In response, we will present a novel statistical framework to mitigate this issue.